Skip to content
/ AutoMS Public

Deep Denoising Autoencoder-assisted Continuous Scoring of Peak Quality in High-Resolution LC−MS Data

License

Notifications You must be signed in to change notification settings

hcji/AutoMS

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

31 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

AutoMS GitHub

AutoMS is a peak picking and quality estimation tool for LC-MS data processing, which utilizes a denoising autoencoder (DAE) to find the common characteristics of chromatographic peaks, predict the noise-free peaks from input ROIs, and evaluate peak qualities. AutoMS integrates HPIC for ROI extraction in order to accept raw data directly and output quantitative results. It also supports peak lists obtained from other tools with little adjustment.

Depends:

Anaconda

Installation

  1. Create conda environment and activate:

           conda create -n automs python=3.8
           conda activate automs
    
  2. Clone the repository and enter:

           git clone https://github.com/hcji/AutoMS.git
           cd ./AutoMS
    
  3. Install dependency:

           pip install -r requirements.txt
    
  4. Run python

           python
    

Usage (with HPIC)

Please download the demo data, unzip to the data folder and run the following codes.

    from AutoMS.automs import AutoMS

    file = 'data/600mix_pos.mzML'
    peaks = AutoMS(file, min_intensity = 5000)

Usage (with XCMS/MZMine)

Please download the demo data, unzip to the data folder and run the following codes.

    # need install rpy2: pip install rpy2,
    # and set R_HOME first (R version >= 3.4.1 and <= 4.1.1).
    
    import pandas as pd
    from AutoMS.automs_external import AutoMS_External
    
    file = 'data/600mix_pos.mzML'
    
    ## Need reset the column name of xcms/mzmine output refer to demo file.
    peaks = pd.read_csv('data/xcms_mzmine_input_demo.csv')
    peaks = AutoMS_External(file, peaks)

Paper Reproduce

All the results in the paper can be reproduced by the scripts in the notebook dir. Please download the dataset in the figshare, and unzip them to the data dir.

Contact

Ji Hongchao
E-mail: ji.hongchao@foxmail.com

WeChat public account: Chemocoder

About

Deep Denoising Autoencoder-assisted Continuous Scoring of Peak Quality in High-Resolution LC−MS Data

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages